Adoptable Coastal Remote Sensing Using Wave-field Observations: Instruments, Techniques and Application

The aim of this study is to innovate wave-based depth-inversion towards smarter and faster algorithms to be used with various remote sensing instruments for broad community use. Wave-based depth inversion describes a branch of coastal remote sensing, which uses video recordings of a wave-field to derive depths and thereby create digital maps of coastal bathymetries. The technique utilizes the fact that waves react to the underlying bathymetry by changing their length and celerity, respectively getting shorter and slower as the water depth gets shallower. Waves may also change their direction due to refraction. Depth inversion techniques using surface wave patterns can handle clear and turbid waters and thereby a variety of global coastal environments. The idea to use observed wave characteristics as a proxy for the underlying bathymetry already came up during the time of the second world war, when the aim was to acquire bathymetry information for military landing operations. Starting around the 1980’s, the idea received more attention among the coastal engineering community, as increased computational power enabled easier analysis of wave-field recordings through spectral decompositions. Since then, different depth inversion algorithms (DIA) have been developed in pursuit of getting increasingly accurate bathymetry maps. Besides estimating depths, some DIAs also incorporate functionalities to map wave propagation directions and wave celerity, and even near-surface currents from wave-field video. While the video recording instruments resemble the hardware, DIAs resemble the software needed for wavebased coastal remote sensing (WCRS). Yet, WCRS is a specialistic branch within the coastal engineering and -user community. The technique typically requires a certain amount of user-expertise and it has mostly been applied in research settings. While data can be retrieved on kilometre scale with XBand-radars and cameras, it was historically difficult to scale up WCRS to entire coasts, which was a reason to discontinue its application in the Netherlands. Besides land-based instruments (i.e., XBand-radars, fixed camera stations) in the meantime also airborne UAVs, and space-borne satellites can be used to record a wave field, making WCRS more flexible and scalable. These recording instruments have also become more accessible. Moreover, DIAs – the software required to analyse the wave recordings – can be used interchangeably on data of these different instruments. This means that WCRS becomes potentially attractive to a broad user-community of coastal managers, the industry and the coast guard. However, DIAs still restrict broad usage of WCRS: while an important step has been taken in the open accessibility of DIAs, much is still to be gained in their handling and computational speed. This study aims to improve upon that, by building towards operational, self-adaptive and intelligent algorithms, which can provide maps of depth, near-surface currents and wave hydrodynamics on-the-fly. For this purpose, video data from a variety of instruments (fixed camera station, UAV, XBandradar, satellite) on different spatial scales 𝑂(100 m2,1 km2,10 km2,100 km2) and field-sites around the world (Netherlands, UK, USA, Australia, France) are analysed. Combining rapid processing capabilities with a broad applicability this study forms a stepping stone for a potentially broad WCRS user community. The analyses are presented going from land-based to air-borne to space-borne WCRS. This is done in three stages from (1) applying an operational DIA on XBand radar data, to (2) applying an on-the-fly DIA on camera and UAV data, to finally (3) applying a DIA on temporally sparse satellite data.First, a DIA named XMFit (X-Band Matlab Fitting) is introduced, which is robust, accurate and fast enough for operational use. This is achieved through an iterative procedure that selects the best result among a series of depth and near-surface current estimates. For this study, video data from XBand-radars are analysed. Focusing on depth estimates, XMFit is validated for two case studies in the Netherlands: (1) the “Sand Engine”, a beach mega nourishment at a uniform open coast, and (2) the tidal inlet of the Dutch Wadden Sea island Ameland, characterizing a more complex coast. Considering both sites, the algorithm performance is characterized by a spatially averaged depth bias of −0.9 m at the Sand Engine (corresponding to an 18 h snapshot of the field site) and a time-varying bias of approximately −2–0 m at the Ameland Inlet (corresponding to a one-year time evolution with varying hydrodynamic conditions). When compared to in-situ depth surveys the accuracy is lower, but the time resolution higher. Dutch in-situ surveys typically occur annually, while depth estimates from the Ameland tidal inlet are produced every 50 min by an operational system using a navigational X-Band radar. It enables to monitor the placement of a 5 Mm3 ebb-tidal delta nourishment – a pilot measure for coastal management. Volumetric changes in the nourishment area over the year 2018, occurring at 7 km distance from the radar, are estimated with an error of 7 %. Depth errors statistically correlate with the direction and magnitude of simultaneous near-surface current estimates. Additional experiments on Sand Engine data demonstrate that depth errors may be significantly reduced using an alternative spectral approach and/or by using a Kalman filter. Having demonstrated the potential of DIAs for operational application, the next step is to design an algorithm that can self-adapt to video from any field-site and can process it on-the-fly. To do so, a DIA is designed whose code architecture for the first time includes the Dynamic Mode Decomposition (DMD) to reduce the data complexity of wavefield video. The DMD is paired with loss-functions to handle spectral noise, and a novel spectral storage system and Kalman filter to achieve fast converging measurements. The algorithm is showcased for videos from ARGUS stations and drones recorded at fieldsites in the USA, UK, Netherlands, and Australia. The performance with respect to mapping bathymetry is validated using ground truth data. It is demonstrated that merely 32 s of video footage is needed for a first mapping update with average depth errors of 0.9–2.6 m. These further reduce to 0.5–1.4 m as the videos continue and more mapping updates are returned. Simultaneously, coherent maps for wave direction and -celerity are achieved as well as maps of local near-surface currents. The algorithm is capable of mapping the coastal parameters on-the-fly and thereby offers analysis of video feeds, such as from drones or operational camera installations. Hence, the innovative application of analysis techniques like the DMD enables both accurate and unprecedentedly fast coastal reconnaissance. With a skilled, intelligent DIA at hand, the question remains whether it can also be used on satellite imagery, as that would further broaden the application range. DIAs commonly analyse video from shore-based camera stations, UAVs or XBandradars with durations of minutes and at framerates of 1–2 fps to find relevant wave frequencies. However, these requirements are typically not met by raw, temporally sparse satellite imagery. To overcome this problem a preprocessing step is utilized. Here, a sequence of 12 images of Capbreton, France, collected over a period of ∼1.5 min at a framerate of 1/8 fps by the Pleiades satellite, is augmented to a pseudo-video with a framerate of 1 fps. For this purpose a recently developed method is used, which considers spatial pathways of propagating waves for temporal video reconstruction. The resulting video is subsequently processed with the self-adaptive DIA. The combination of image augmentation with a frequency-based depth inversion method shows potential for broad application to temporally sparse satellite imagery and thereby aids in the effort towards broad usage of WCRS for mapping coastal bathymetry data around the globe. By improving DIAs and their application to different instruments, this study has helped to increase the technological readiness of WCRS and its potential to be adopted by end-users. It was shown that WCRS can be performed on wave field records of land-based, airborne and space-born instruments and therewith on scales ranging from 𝑂(100 m2)(fixed camera) to 𝑂(100 km2)(X-band radar,satellite). The cost of WCRS is minor, as existing navigational X-band radars can be used, affordable UAVs and cameras, and accessible satellite data. X-band radars can operationally monitor complex coastal environments and recognize morphological trends, UAVs and cameras can be used for fast lean-and-mean mapping of coastal bathymetry, and by estimating depths from satellite imagery valuable data can be collected in otherwise data-poor environments. Yet, further steps should be taken in the accessibility, multifunctionality, quality, robustness and user-friendliness of WCRS. The key takeaway for effective WCRS monitoring is that future developments should strive towards integrated, self-adaptive software, which gives prompt visual response and requires little user-expertise. These measures reduce the difficulty to learn WCRS, increase its compatibility with data from different instruments (Xband-radars, cameras, UAVs, satellites) and thereby enable relatively easy coastal measurements. As a consequence WCRS becomes more adoptable by the coastal remote sensing community. With the exponential growth of data volumes worldwide, future data clouds may facilitate storage and offer future perspectives for online integration of data with numerical models and modern data science techniques like neural networks. This may create new possibilities for understanding system dynamics and thereby further aid decision makers in coastal management, the industry and the coast guard.Coastal Engineerin